Computerized Pattern Recognition Applied to Battery Testing.
Abstract
The primary goal of this work has been to develop non-destructive testing methods as a screening procedure for batteries to predict lifetime and identify probably failure mechanisms. A secondary goal has been to develop criteria for predicting imminent failure from battery performance data. We believe that these goals can be met by the application of computerized pattern recognition to the evaluation of measureable features. In the studies supported by this ONR contract, pattern recognition techniques have been used to analyze data collected previously in established battery testing programs. Because of the ability to evaluate multivariate relationships it was possible to identify previously unobservable multi-dimensional correlations between test data obtained very early in a battery's life-time and its life expectancy and/or failure mechanism. It was the objective of these initial studies to identify the most meaningful types of measurements for lifetime prediction and to establish the feasibility of lifetime prediction. The second function to be provided by pattern recognition methods is to evaluate test data generated subsequently by the new short-term screening tests developed. This will involve collaborative efforts with other laboratories to generate such a data base. Probably several iterations in this procedure will be required before an optimum set of short-term measurements is identified. Such studies are prescribed for future work. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1980
- Accession Number
- ADA099410
Entities
People
- Sam P. Perone
Organizations
- Purdue University